Effect of sodium–glucose cotransporter 2 inhibitors on heart failure readmission in older patients with diabetes: A Japanese claims‐based cohort study

ABSTRACT

Aims

To evaluate the effect of sodium–glucose cotransporter 2 inhibitor (SGLT2i) administration at hospital discharge on heart failure (HF) readmission among older patients (≥75 years) with HF and diabetes and to compare the effectiveness of four SGLT2is, namely, ipragliflozin, empagliflozin, canagliflozin, and dapagliflozin, with a focus on direct head‐to‐head comparisons among individual SGLT2is.

Materials and Methods

We conducted a retrospective cohort study using the Medical Data Vision claims database. We included 66,895 patients aged ≥75 years who were emergently hospitalized for HF with comorbid diabetes between 2018 and 2022. Patients were categorized according to SGLT2i administration at discharge. Propensity score matching (PSM) was applied to adjust for baseline characteristics. The primary outcome was HF readmission, assessed using Kaplan–Meier analysis and Cox proportional hazards models. Inverse probability of treatment weighting (IPTW) was used to compare the four SGLT2is.

Results

After PSM, SGLT2i users showed a significantly lower risk of HF readmission than nonusers (hazard ratio [HR] 0.91, 95% confidence interval [CI]: 0.85–0.97, P = 0.003). Sensitivity analysis censoring death and treatment discontinuation or initiation produced consistent results (HR 0.79, 95% CI: 0.74–0.85, P < 0.001). An IPTW‐based comparison of the four agents revealed no significant differences, suggesting similar class‐wide effectiveness.

Conclusions

Among older patients with HF and diabetes, SGLT2i administration was associated with reduced HF readmission risk, with no meaningful differences among the four agents. These findings support a class‐wide effect of SGLT2i and highlight the importance of their appropriate initiation and continuation in older adults.

SGLT2 inhibitor use at discharge was associated with a reduced risk of heart failure readmission in elderly patients with diabetes, with no significant differences among individual agents.

INTRODUCTION

Heart failure (HF) becomes increasingly prevalent with age and frequently requires hospitalization in older adults ¹ , ² , ³ . More than 25% of patients are readmitted within 1 year of discharge, particularly among older adults ⁴ , ⁵ , ⁶ , ⁷ . Diabetes mellitus is a major risk factor for HF and is associated with poorer prognosis ⁸ , ⁹ , ¹⁰ , ¹¹ .

Sodium–glucose cotransporter 2 inhibitors (SGLT2is) have been approved as antidiabetic agents and have recently been shown to be effective against HF. Large randomized controlled trials have demonstrated that dapagliflozin (DGZ) and empagliflozin (EGZ) reduce the risk of HF hospitalization, leading to the expansion of HF indications for these agents in Japan ¹² , ¹³ , ¹⁴ . However, real‐world data directly comparing the effects of individual SGLT2is on HF readmission remain limited. Although SGLT2is share class‐wide effects ¹⁵ , ¹⁶ , they differ in molecular structure and pharmacokinetics ¹⁷ , ¹⁸ , making it important to clarify whether drug‐specific differences exist in preventing HF readmission.

Large‐scale cohort studies ¹⁹ have demonstrated that SGLT2is reduce HF readmission among patients with diabetes. However, these studies have primarily focused on comparisons with other classes of antidiabetic agents, and direct head‐to‐head comparisons among individual SGLT2is in very elderly populations remain scant. In this study, we addressed this evidence gap by analyzing a real‐world cohort of patients aged ≥75 years. We examined the association between SGLT2i use and HF readmission using propensity score matching (PSM) and further evaluated differences in effectiveness among the four major SGLT2is by applying inverse probability of treatment weighting (IPTW). Determining whether clinically meaningful differences exist among individual SGLT2is—or whether their effects represent a class‐wide benefit—may inform therapeutic decision‐making and help optimize HF management in elderly patients.

MATERIALS AND METHODS

Data source

This retrospective cohort study used anonymized data from the Medical Data Vision (MDV) claims database (Medical Data Vision Co., Ltd., Tokyo, Japan). The MDV database includes information related to diagnosis, prescriptions, procedures, laboratory test results, surgeries, and hospitalization details collected from more than 540 acute‐care hospitals across Japan, covering approximately 50 million patients as of February 2024. Diagnostic information in the database is coded using the International Classification of Diseases, 10th Revision (ICD‐10).

In this study, we used anonymized data of patients aged ≥75 years who were emergently hospitalized between September 1, 2018 and November 30, 2023. In the MDV database, hospitalizations are classified as ‘scheduled/other’ or ‘emergency medical admission’, with the latter defined as an emergency hospitalization. Longitudinal medical records can be traced back to April 2008.

Ethics approval and informed consent

This study was conducted in accordance with the Declaration of Helsinki and the Ethical Guidelines for Medical and Biological Research Involving Human Subjects in Japan. The study protocol was approved by the Ethics Committee of Hokkaido University of Science, Hokkaido, Japan (approval No.: 25‐17). Given that this study only used fully anonymized secondary data, individual patients could not be identified, and the requirement for written informed consent was waived ²⁰ .

Patients

We identified 1,831,331 emergency hospitalizations among patients aged ≥75 years between September 1, 2018 and November 30, 2022. Although data were available until November 30, 2023, admissions resulting in discharge on or after December 1, 2022 were excluded to ensure at least 1 year of follow‐up after hospital discharge. Among these hospitalizations, 145,861 admissions were due to HF (ICD‐10: I50), and 84,659 of these belonged to patients diagnosed with type 2 diabetes (ICD‐10: E11). We excluded patients who had been hospitalized for HF within 1 year prior to the index emergency admission to avoid including patients with recurrent or chronically unstable HF, which could confound the assessment of postdischarge readmission risk. Patients who died during the index hospitalization were also excluded. The final study population consisted of 66,895 patients (Figure 1).

Figure 1.

Cohort selection flowchart. Flowchart illustrating the process of cohort selection for this study. Patients aged ≥75 years with a history of DM who were emergently hospitalized for HF were identified from the MDV database in Japan. Patients with a prior hospitalization for HF within the past year or those who died during the index hospitalization were excluded. To ensure a minimum follow‐up period of 1 year, admissions resulting in discharge on or after December 1, 2022 were excluded. The final cohort was divided based on SGLT2i use at discharge. Diagnosis codes were identified using the International Classification of Diseases, 10th Revision (ICD‐10). CGZ, canagliflozin; DGZ, dapagliflozin; DM, type 2 diabetes mellitus; E11 = type 2 diabetes mellitus; EGZ, empagliflozin; HF, heart failure; I50 = heart failure; IGZ, ipragliflozin; IPTW, inverse probability of treatment weighting; LGZ, luseogliflozin; MDV, Medical Data Vision; PSM, propensity score matching; SGLT2i, sodium–glucose cotransporter 2 inhibitor; TGZ, tofogliflozin.

Patients were categorized into an SGLT2i user group (n = 8,360) and a nonuser group (n = 58,535) according to SGLT2i administration at discharge. A discharge prescription was defined as a prescription recorded from 3 days before to the day of discharge, where the days supplied extended beyond the discharge date, indicating continued administration after discharge. PSM was used to compare SGLT2i users and nonusers, and IPTW was performed to compare individual SGLT2i drugs. Considering that a small sample size can compromise estimation stability ²¹ , tofogliflozin (TGZ) and luseogliflozin (LGZ) were excluded from the IPTW analyses. Accordingly, 8,183 patients prescribed ipragliflozin (IGZ), EGZ, canagliflozin (CGZ), or DGZ at discharge were included in the drug‐to‐drug comparison. The propensity score was estimated using all baseline demographic, functional, comorbidity, and medication variables described in section ‘Patients’.

Variables

Administration of SGLT2is at discharge was defined as the exposure, HF readmission as the primary outcome, and all baseline variables were considered potential confounders. Baseline characteristics were recorded at admission. Demographic and functional variables included age, sex, cognitive function status ²² , long‐term care level, and level of consciousness (Japan Coma Scale) ²³ . Cognitive status was classified according to the Ministry of Health, Labour and Welfare categories for the degree of independence in daily living for older adults with dementia (ranks I–IV and M). Patients without dementia were categorized as ‘No impairment’. Long‐term care level was classified as ‘Independent’, ‘Support needed’, ‘Long‐term care’, or ‘Pending’. Comorbidities included atrial fibrillation, ischemic heart disease, valvular disease, stroke or transient ischemic attack, chronic kidney disease (CKD), diabetic nephropathy (DN), dyslipidemia, obesity, chronic bronchitis or chronic obstructive pulmonary disease, dementia, and malignancy. CKD was further categorized as stages G1–3, stages G4–5 (advanced CKD, including dialysis), or unspecified stage based on diagnosis codes. Data on medication use at admission were extracted for HF‐related drugs, including angiotensin‐converting enzyme inhibitors, angiotensin II receptor blockers, angiotensin receptor–neprilysin inhibitors (ARNI), catecholamines, coronary vasodilators, digitalis, hyperpolarization‐activated cyclic nucleotide–gated channel inhibitors, potassium channel openers, potassium‐sparing diuretics, loop diuretics, mineralocorticoid receptor blockers, noncatecholamine inotropes, soluble guanylate cyclase stimulators, thiazide diuretics, V2 receptor antagonists, and beta‐blockers. Antidiabetic medications included biguanides, dipeptidyl peptidase‐4 inhibitors (DPP‐4i), glinides, glucagon‐like peptide‐1 receptor agonists, insulin, SGLT2is, sulfonylureas, thiazolidinediones, and alpha‐glucosidase inhibitors. ICD‐10 and receipt codes used to define comorbidities and medications are provided in Tables S1– S3. Items lacking diagnostic or prescription records were considered ‘not recorded’. No sample size calculations were performed because all eligible cases in the MDV database were included.

Statistical analyses

For the PSM analysis, the cumulative incidence of HF readmission was estimated using Kaplan–Meier methods, and intergroup differences were evaluated using the log‐rank and Gehan–Breslow–Wilcoxon tests. For drug‐to‐drug comparisons, Cox proportional hazards models incorporating IPTW with stabilized weights were used ²⁴ . PSM and IPTW were applied to mitigate potential confounding bias. The follow‐up period extended from the index discharge date until HF readmission, death, loss to follow‐up, or the end of available data (November 30, 2023).

For the PSM analysis, the primary censoring criteria were death and loss to follow‐up:

1. Death was defined as any hospitalization after discharge in which the recorded discharge outcome was ‘death’.

2. Loss to follow‐up was defined as the date of the last available inpatient or outpatient record, with censoring applied at that final documented encounter in the MDV database.

In the corresponding sensitivity analysis, the censoring criteria included all events in the primary analysis plus two additional events, for a total of four censoring criteria:

• 3Discontinuation of SGLT2i therapy in the user group, defined as cases in which the last prescription date plus the days supplied did not extend beyond the final database date (November 30, 2023).
• 4Initiation of SGLT2i therapy in the nonuser group, defined as the first SGLT2i prescription recorded after discharge among patients not initially prescribed an SGLT2i.

For the IPTW analyses comparing individual SGLT2i drugs, the same censoring framework used in the PSM analyses was applied, including both the primary and sensitivity censoring criteria.

Standardized differences (std diff) were used to evaluate covariate balance, with a threshold of <0.10 indicating acceptable balance ²⁵ . Statistical analyses were performed using JMP® Student Edition 18 (SAS Institute Inc., Cary, NC, USA) and R version 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Two‐sided P‐values <0.05 were considered statistically significant.

RESULTS

Association between SGLT2i administration and HF readmission

Covariate balance after PSM

Of the 66,895 patients comprising the study cohort, 8,360 were prescribed an SGLT2i at discharge, and 58,535 were not. After 1:1 PSM, 7,430 patients were retained in each group. Baseline characteristics after matching are listed in Table 1. The Std diff for all variables related to demographic characteristics, functional status, comorbidities, HF‐related medications, and antidiabetic agents was <0.10, indicating good balance between groups. Among the SGLT2i users, 647 patients (8.7%) were taking an SGLT2i at admission and continued the therapy at discharge. Conversely, 533 patients (7.2%) in the nonuser group had been taking an SGLT2i prior to admission but discontinued it at discharge.

Table 1.

Baseline characteristics of patients after propensity score matching

	SGLT2i users (n = 7,430)	Nonusers (n = 7,430)	|std diff|
Demographics and functional status
Age (years)	82.9 ± 5.2	82.7 ± 5.6	0.015
Sex (Male/Female)	4,294/3,136	4,352/3,078	0.016
Cognitive function status (No impairment/I/II/III/IV/M)	5,313/957/584/443/107/26	5,447/892/568/406/97/20	0.021
Long‐term care level (Independent/Support needed/Long‐term care/Pending)	5,114/727/1,353/236	5,185/714/1,330/201	0.016
Level of consciousness (JCS) (I/II/III/Alert)	982/110/46/6,292	942/102/44/6,342	0.012
Comorbidities
AF	3,988	3,966	0.006
IHD	5,689	5,748	0.019
Valvular disease	2,195	2,152	0.013
Stroke/TIA	1,730	1,642	0.028
CKD (G1–3/G4–5/unspecified)	309/494/1,818	291/491/1,837	0.005
DN	4,707	4,749	0.012
Dyslipidemia	4,829	4,829	0
Obesity	55	47	0.013
Chronic bronchitis/COPD	511	447	0.035
Dementia	536	485	0.027
Malignancy	3,477	3,376	0.027
HF‐related medications
ACEi	274	243	0.023
ARB	136	129	0.007
ARNI	130	107	0.025
Catecholamines	7	6	0.005
Coronary vasodilators	8	6	0.009
Digitalis	95	85	0.012
HCNi	8	8	0
K‐channel openers	202	198	0.003
K‐sparing diuretics	602	525	0.034
Loop diuretics	1,592	1,448	0.057
MRB	118	117	0.028
Non‐catechol inotropes	116	104	0.015
sGC stimulators	2	1	0.023
Thiazide diuretics	138	142	0.018
V2RA	484	457	0.022
β‐blockers	1,400	1,255	0.054
Antidiabetic medications
Biguanides	284	253	0.022
DPP‐4i	608	554	0.027
Glinides	167	150	0.016
GLP‐1RA	75	58	0.024
Insulin	405	371	0.021
SGLT2i	647	533	0.057
SU	245	245	0
TZD	73	72	0.001
α‐GI	228	213	0.012

Continuous variables are presented as mean ± standard deviation, and categorical variables as counts. Absolute std diff are reported for each variable to assess covariate balance after propensity score matching. Absolute standardized difference of <0.1 was considered indicative of good balance between groups. All baseline variables were assessed at admission, reflecting patient status prior to the index hospitalization.

ACEi, angiotensin‐converting enzyme inhibitor; AF, atrial fibrillation or flutter; ARB, angiotensin II receptor blocker; ARNI, angiotensin receptor–neprilysin inhibitor; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; DN, diabetic nephropathy; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; GLP‐1RA, glucagon‐like peptide‐1 receptor agonist; HCNi, HCN channel inhibitor; IHD, ischemic heart disease; JCS, Japan Coma Scale; MRB, mineralocorticoid receptor blocker; sGC, soluble guanylate cyclase; SGLT2i, sodium–glucose cotransporter 2 inhibitor; std diff, standardized difference; SU, sulfonylurea; TIA, transient ischemic attack; TZD, thiazolidinedione; V2RA, vasopressin V2 receptor antagonist; α‐GI, alpha‐glucosidase inhibitor.

Kaplan–Meier analysis of HF readmission (after PSM)

During follow‐up, 1,659 and 2,001 HF readmissions occurred in the SGLT2i group and the nonuser group, respectively. Kaplan–Meier curves are shown in Figure 2a. In the primary analysis (censoring at death), SGLT2i users had significantly lower rates of HF readmission than did nonusers (log‐rank P = 0.003; Gehan–Breslow–Wilcoxon P < 0.001). The hazard ratio (HR) for 1‐year HF readmission calculated using the Cox proportional hazards model was 0.91 (95% confidence interval [CI]: 0.85–0.97), indicating that SGLT2i administration was associated with a significantly reduced risk of readmission.

Figure 2.

Kaplan–Meier curves for 5‐year cumulative readmission rates among older patients hospitalized for HF, stratified by SGLT2i administration at discharge. Panel (a) shows the primary analysis, in which patients were censored at death but not at treatment changes (i.e., discontinuation or initiation of SGLT2i). Panel (b) presents the sensitivity analysis, with censoring at both treatment changes and death. In both analyses, SGLT2i users showed significantly lower readmission rates than nonusers. (a) ^†log‐rank test, P = 0.003, HR = 0.91 (95% CI: 0.85–0.97); ^‡Gehan–Breslow–Wilcoxon test, P < 0.001. (b) ^†log‐rank test, P < 0.001, HR = 0.79 (95% CI: 0.74–0.85); ^‡Gehan–Breslow–Wilcoxon test, P < 0.001. CI, confidence interval; HF, heart failure; HR, hazard ratio; SGLT2i, sodium–glucose cotransporter 2 inhibitor.

In the sensitivity analysis, which incorporated treatment discontinuation or initiation as censoring events—the reduction in readmission risk with SGLT2i administration remained significant (log‐rank P < 0.001; Gehan–Breslow–Wilcoxon P < 0.001), with an HR of 0.79 (95% CI: 0.74–0.85), as shown in Figure 2b.

Furthermore, to evaluate the robustness of these findings, we conducted a new‐user sensitivity analysis restricted to patients who initiated SGLT2i therapy at discharge, excluding those with prior SGLT2i use before the index hospitalization. After excluding these patients, PSM was re‐performed, and the baseline characteristics of the matched cohort are presented in Table S4. Covariate balance was satisfactory across all variables. A significantly lower risk of HF readmission was consistently observed in both the primary and sensitivity analyses (Figure S1).

Comparison among SGLT2i agents

Covariate balance after IPTW

After excluding TGZ (n = 112) and LGZ (n = 65) for small sample sizes, IPTW was performed to compare the other four SGLT2i drugs—IGZ (n = 356), EGZ (n = 3,345), CGZ (n = 520), and DGZ (n = 3,962).

Before IPTW, the Std diff exceeded 0.10 for five variables (long‐term care level, DN, ARNI, DPP‐4i administration, and SGLT2i administration). After IPTW, all Std diff were <0.10 (Figure 3). The mean Std diff decreased from 0.066 to 0.020, and the maximum value decreased from 0.316 to 0.050. Results involving TGZ and LGZ (six‐drug comparison) are shown in Figure S2.

Figure 3.

Standardized differences before and after IPTW (TGZ and LGZ excluded). This Love plot displays the absolute std diff in covariates between SGLT2i users and nonusers, before (●) and after (■) applying IPTW. Covariates are categorized into four domains: demographics and functional status, prevalence of comorbidities, HF‐related medications, and antidiabetic medications. The vertical dashed line at 0.1 indicates the threshold for acceptable covariate balance. After IPTW, all covariates showed std diff <0.1, indicating good balance between groups. ACEi, angiotensin‐converting enzyme inhibitor; AF, atrial fibrillation or flutter; ARB, angiotensin II receptor blocker; ARNI, angiotensin receptor–neprilysin inhibitor; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; DN, diabetic nephropathy; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; GLP‐1RA, glucagon‐like peptide‐1 receptor agonist; HCNi, HCN channel inhibitor; HF, heart failure; IHD, ischemic heart disease; IPTW, inverse probability of treatment weighting; JCS, Japan Coma Scale; LGZ, luseogliflozin; MRB, mineralocorticoid receptor blocker; sGC, soluble guanylate cyclase; SGLT2i, sodium–glucose cotransporter 2 inhibitor; std diff, standardized difference; SU, sulfonylurea; TGZ, tofogliflozin; TIA, transient ischemic attack; TZD, thiazolidinedione; V2RA, vasopressin V2 receptor antagonist; α‐GI, alpha‐glucosidase inhibitor.

IPTW‐adjusted cox models for HF readmission

Among the four SGLT2i drugs, DGZ had the largest sample size and was used as the reference agent in drug‐to‐drug comparisons. Using IPTW‐adjusted Cox proportional hazards models, we compared the risk of HF readmission within 1 year after discharge for IGZ, EGZ, and CGZ relative to DGZ (Figure 4). In the primary analysis, none of the agents significantly differed from DGZ (IGZ: HR 0.85, 95% CI: 0.63–1.13, P = 0.265; EGZ: HR 1.03, 95% CI: 0.93–1.13, P = 0.581; CGZ: HR 1.01, 95% CI: 0.81–1.25, P = 0.925). The sensitivity analysis yielded consistent results, with no significant differences observed (IGZ: HR 0.81, 95% CI: 0.59–1.13, P = 0.216; EGZ: HR 1.03, 95% CI: 0.92–1.14, P = 0.627; CGZ: HR 0.95, 95% CI: 0.78–1.22, P = 0.828). Results involving TGZ and LGZ (six‐drug comparison) are shown in Figure S3.

Figure 4.

Association between selected individual SGLT2is and risk of 1‐year HF readmission. Forest plot showing hazard ratios and 95% confidence intervals for 1‐year HF readmission associated with three selected SGLT2is, using DGZ as the reference. IPTW was applied to adjust for baseline covariates across treatment groups. No statistically significant differences in HF readmission risk were observed among the agents, as all confidence intervals crossed unity. (a) Primary analysis: censored at death only. (b) Sensitivity analysis: censored at both treatment change and death. All comparisons showed no statistically significant differences among the agents. CGZ, canagliflozin; CI, confidence interval; DGZ, dapagliflozin; EGZ, empagliflozin; HF, heart failure; IGZ, ipragliflozin; IPTW, inverse probability of treatment weighting; SGLT2i, sodium–glucose cotransporter 2 inhibitor.

DISCUSSION

In this large‐scale real‐world cohort of patients aged ≥75 years with HF and comorbid diabetes, SGLT2i administration at discharge was associated with a markedly lower risk of HF readmission than nonuse. Furthermore, no statistically significant differences were observed among the four SGLT2i drugs—IGZ, EGZ, CGZ, and DGZ—after adjustment with IPTW. These findings suggest that SGLT2i administration reduces HF readmission. The direct comparison of multiple SGLT2i agents in this very old population is a notable strength of this study. These findings were further supported by a new‐user sensitivity analysis, which yielded results consistent with the primary analysis (Figure S1).

Previous studies have similarly shown a protective effect of SGLT2i therapy on HF‐related outcomes among patients with diabetes. In a nationwide claims‐based cohort study using the National Database of Health Insurance Claims and Specific Health Checkups of Japan, Nakai et al. ¹⁹ reported that SGLT2i recipients had a substantially lower risk of HF readmission within 1 year than those receiving DPP‐4i (HR 0.52, 95% CI: 0.45–0.61). Studies not limited to diabetes populations, including the Optimal Treatment for Old‐Aged Patients with a Safe and Effective Approach in the Setouchi Region study ²⁶ and Reduction of Rehospitalization with SGLT2 Inhibitors Efficacy Study in Heart Failure, Setouchi Region study ²⁷ , have also shown that treatment with SGLT2is—primarily DGZ and EGZ—reduced composite endpoints of HF readmission and cardiovascular mortality among patients aged ≥75 years. Our study corroborates these findings but also extends them by evaluating outcomes for up to 5 years, providing additional evidence regarding the sustained, long‐term effectiveness of SGLT2is in very old individuals. The divergence of Kaplan–Meier curves further indicates a persistent benefit over time. Although these findings are broadly consistent with previous studies, the magnitude of the estimated effect differed between the primary and sensitivity analyses in the present study. In the sensitivity analysis, in which patients were censored at treatment discontinuation or initiation, a smaller HR was observed compared with the primary analysis (HR 0.79 vs 0.91). This difference may be attributable to the exclusion of HF readmissions associated with high‐risk events occurring after treatment discontinuation. These findings suggest that continued SGLT2i therapy is associated with greater treatment benefit, whereas treatment discontinuation itself may reflect clinical deterioration and poorer prognosis. In addition, as shown in Figure S4, the proportion of patients receiving SGLT2is increased over time, particularly after 2020, reflecting temporal changes in prescribing patterns following the expansion of HF indications. These temporal trends should be considered when interpreting the results. Despite these considerations, our findings consistently suggest that supporting the continuation of SGLT2i therapy after discharge may have important clinical implications, particularly in elderly patients with HF.

Furthermore, the association between SGLT2i use and a lower risk of HF readmission was generally consistent across the subgroups stratified by cognitive function and level of care dependency, supporting the robustness of our findings (Figure S5).

The mechanisms by which SGLT2i reduce HF readmission extend beyond glucose lowering and are thought to involve natriuresis, plasma volume regulation, and mitigation of glomerular hyperfiltration ²⁶ . Although SGLT2is differ modestly in their pharmacokinetic properties, SGLT2 selectivity, and elimination pathways, they share a common molecular target: the renal proximal tubular SGLT2. In our previous multicenter study of 3,680 patients with diabetes, SGLT2i therapy was associated with improvements in the estimated glomerular filtration rate (eGFR) and aspartate aminotransferase and alanine aminotransferase levels, with no significant differences among the six SGLT2i drugs ¹⁶ . These results highlight the class‐wide reno‐ and hepato‐protective effects of SGLT2is. Consistent with these findings, the present study showed no significant differences among individual SGLT2is in terms of IPTW‐adjusted HF readmission risk, supporting a clinically meaningful class effect in elderly individuals. Importantly, in this study, we performed a comprehensive head‐to‐head comparison among four specific SGLT2is, which represents a key novel aspect of our analysis and offers clinically relevant insights for drug selection in real‐world practice.

Despite their clinical benefits, concerns regarding the safety of SGLT2is in elderly individuals persist, particularly regarding urinary and genital infections, dehydration, and early declines in eGFR ²⁸ . In the SGLT2i study of elderly patients with diabetes ²⁹ , adverse events, including dehydration and infection, were slightly more frequent in older adults but were generally mild and manageable. Meta‐analyses involving older adults have also shown a safety profile comparable to that of younger populations, with severe events remaining rare ³⁰ , ³¹ . Moreover, studies based on the FDA Adverse Event Reporting System have not identified new safety concerns in patients aged ≥75 years ³² . Collectively, these findings suggest that although careful drug initiation and routine monitoring are essential, age alone should not preclude SGLT2i administration. Given the pleiotropic benefits of SGLT2is, especially their cardioprotective, renoprotective, and hepatoprotective effects, it is reasonable to expect that older adults could derive substantial clinical advantages from appropriate SGLT2i administration. The absence of inter‐drug differences in our study suggests that maintaining SGLT2 inhibition, rather than selecting a specific agent, may be the key factor in improving outcomes.

This study has some limitations. First, given that this was a retrospective observational study using a claims database, key clinical indicators reflecting HF severity, such as left ventricular ejection fraction and B‐type natriuretic peptide (BNP)/N‐terminal pro–BNP levels, were unavailable. Therefore, we were unable to perform analyses stratified by HF phenotype (HFrEF vs HFpEF). However, SGLT2is have been reported to reduce HF hospitalization events across the entire ejection fraction spectrum, and thus, the observed benefits in this study, as well as the lack of clear differences between individual agents, may reflect a class effect ¹³ , ³³ . Nevertheless, this aspect could not be directly evaluated within the present study design and the findings should be interpreted with caution. In addition, although frailty and sarcopenia are important prognostic determinants in older patients with HF ³⁴ , ³⁵ , clinically relevant information related to overall health status, medication dose, and medication adherence could not be adequately captured. Second, HF readmission was defined based on ICD‐10 codes, and differences in diagnostic coding practices across institutions may have led to outcome misclassification. Furthermore, patients prescribed SGLT2is at discharge may have had more stable clinical status or received more intensive, guideline‐adherent care, introducing the potential for indication bias. However, we attempted to mitigate these effects by applying propensity score methods and conducting multiple sensitivity analyses. These limitations warrant cautious interpretation, and the findings should be considered hypothesis‐generating rather than causal.

CONCLUSION

This large‐scale real‐world cohort study of patients aged ≥75 years with HF and diabetes showed that SGLT2i administration at hospital discharge was associated with a significantly lower risk of HF readmission, with no meaningful differences observed among the four commonly prescribed SGLT2i drugs, supporting a class‐wide effect. These results highlight the importance of appropriately initiating and maintaining SGLT2i therapy in older adults rather than prioritizing specific agents. Future prospective studies incorporating clinical and frailty‐related measures are required to further validate these findings. The present study provides valuable evidence to guide treatment selection in older patients with HF and diabetes in routine clinical practice.

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: The study protocol was approved by the Ethics Committee of Hokkaido University of Science, Hokkaido, Japan (approval No.: 25‐17).

Informed Consent: Because this study used fully anonymized secondary data obtained from the Medical Data Vision claims database, individual patients could not be identified, and the requirement for written informed consent was waived.

Approval date of Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.

Supporting information

Figure S1. Kaplan–Meier curves for 5‐year cumulative HF readmission rates among elderly patients hospitalized for heart failure, restricted to patients who initiated SGLT2i therapy at discharge.

Figure S2. Std diff before and after IPTW (all SGLT2i included).

Figure S3. Association between individual SGLT2i and risk of 1‐year readmission.

Figure S4. Temporal distribution of discharge periods among SGLT2i users and nonusers.

Figure S5. Subgroup analysis of the association between SGLT2i use and heart failure readmission stratified by cognitive function and level of care dependency.

Table S1. Definitions of comorbidities based on ICD‐10 code.

Table S2. HF‐related medications.

Table S3. Antidiabetic medications.

Table S4. Baseline characteristics of patients after propensity score matching in the new‐user analysis.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.